Osteoarthritis (OA) is the most common rheumatic disease and most prevalent form of arthritis in the United States. Primary goals of its management are to reduce pain, and improve functional ability and quality of life. Patient education and exercise are recommended as two mainstays of treatment. Thus far, there are only a few well-designed studies that utilize education, exercise, or both as an intervention. Most studies have not shown adequate internal validity, have not measured all relevant outcomes, have included only advanced-disease patients and have not examined the long-term benefits of the interventions. We hypothesize that the implementation of a long-term multidimensional intervention incorporating both an exercise and a self-management program in a sample of subjects with early knee OA will result in a significant improvement in physical function, as compared with each of the two programs alone. We aim to test this hypothesis based on a model that assumes the exercise intervention would have both direct and indirect effects (mediated by coping efficacy and affectivity) on physical function. The addition of a self-management program would have an additive effect on physical function, through its effect on coping efficacy; or, alternatively a multiplicative effect moderating the effect of exercise on coping efficacy and affectivity. In conclusion, given that multiple factors are involved in this dynamic model, the use of a multidimensional program, combining exercise training and self-management, should result in a stronger effect on long-term maintenance of physical function and prevention of disability. For this purpose we will randomize 300 early OA patients to: 1) a 24-month intervention combining a multifaceted exercise training plus a self-management program; 2) an exercise program; or 3) a self-management program.